03 · What You Need to Know
A changing plan is manageable when the change itself is managed
Research protocols are sometimes imagined as contracts with the future: once written, nothing should move.
That is not how research always works.
Piloting exists partly because procedures may need refinement. Qualitative methodologies may incorporate iterative sampling or analysis. Recruitment assumptions can prove wrong. External conditions can change. Equipment fails. New information can reveal a risk that was not apparent during planning.
At the same time, the opposite position, that researchers can change anything whenever circumstances become inconvenient, is equally problematic.
The useful principle is controlled change: modifications should have a reason, consequences should be assessed, applicable governance requirements should be followed, and important departures from the original plan should remain visible.
First determine what actually changed
“The research plan changed” is too broad to guide a response.
Identify the specific decision or procedure that is different.
For example:
- one recruitment site is no longer available;
- the target population has been narrowed;
- an eligibility criterion has changed;
- a survey item has been removed;
- the interview guide includes a new topic;
- the data collection period has been extended;
- the primary outcome has changed;
- a planned statistical model has been replaced;
- an additional data source has been introduced; or
- the retention period for identifiable information has changed.
Specificity matters because each kind of change has different consequences.
Then ask why the change is being made
The rationale can help distinguish a necessary adaptation from a questionable post hoc decision.
Possible reasons include:
| Reason for change |
Example |
Main question to examine |
| Feasibility |
Recruitment is substantially slower than expected. |
Can the study remain scientifically adequate under a revised recruitment strategy or scope? |
| Pilot or pretest evidence |
Participants consistently misunderstand an item. |
What should be corrected before the main study? |
| Ethical concern |
A procedure creates greater burden or distress than anticipated. |
What must change to protect participants, and what review is required? |
| Technical failure |
Equipment or software becomes unavailable. |
Can an alternative produce comparable and appropriate evidence? |
| Data problem |
An existing dataset lacks a required variable. |
Does the revised evidence still answer the original question? |
| Methodological discovery |
A planned analytical assumption is untenable. |
What alternative analysis is defensible, and how should the change be reported? |
| External change |
A participating organization withdraws. |
What does the loss change about access, sampling, scope, and feasibility? |
| Result-dependent choice |
A researcher wants to change the primary outcome after seeing results. |
Would the observed results be influencing a decision that was supposed to be prospective? |
The final category deserves particular caution. A change made because new methodological information emerged is different from a change made because one analysis produced a more attractive result.
Classify the change by consequence, not by how small it looks
A one-word revision can sometimes matter more than a five-page procedural update.
Changing an internal meeting schedule may have almost no scientific consequence. Changing one word in an eligibility criterion could alter who enters the study.
Rather than classifying modifications according to how much text changes, ask what the modification affects.
Operational change
Primarily changes how work is organized without materially altering participants, evidence, scientific interpretation, or applicable requirements.
Scientific or governed change
Changes what is studied, who participates, what happens to participants, what data are generated, how evidence is interpreted, or an element governed by an approval, agreement, registration, or other requirement.
The boundary is contextual. A change in recruitment location might be merely logistical in one study but substantially alter the population in another.
Trace the change through the rest of the study
Research decisions form a dependency network. Changing one can create several downstream changes.
Suppose you expand recruitment to a second institution because the original site is recruiting too slowly.
That may affect:
- site authorization;
- ethics documentation;
- participant recruitment materials;
- sampling and representativeness;
- data collection logistics;
- site identifiers;
- data-security arrangements;
- analysis if site differences become relevant;
- budget;
- researcher workload; and
- the timeline.
The change is therefore not “add another recruitment site.” It is a change whose consequences need to be propagated across the study.
This is why mapping research dependencies is useful even after planning appears complete.
Check whether the research question still matches the revised study
Some modifications leave the original question intact. Others quietly change what the study can answer.
Suppose your question asks whether an intervention improves academic performance compared with usual practice, but circumstances force you to remove the comparison condition. The resulting study may still be worthwhile, but it may no longer support the original comparative claim.
Similarly, if a required variable is unavailable in a secondary dataset, replacing it with a convenient proxy may alter the construct being investigated.
After a consequential change, ask:
If I conducted the revised study exactly as now planned, would its evidence still answer the original research question?
If not, the question, objectives, claims, or design may also need revision.
Check whether the change alters the population or sample
Changes to recruitment, eligibility, sites, sampling frames, or data sources can alter who or what the findings represent.
Imagine that a study initially planned to recruit students from both public and private universities but ultimately recruits only from one private institution because access elsewhere fails.
The problem is not simply that the sample is smaller. The evidentiary scope has changed.
The revised study may still be valid for a narrower population, but the eventual claims should reflect that narrower evidence.
This is why feasibility adaptations should not be described merely as logistical adjustments when they materially alter the sample or setting.
Check whether earlier and later data remain comparable
A mid-study procedural change creates a particularly important question:
Are observations collected before and after the change still meaningfully comparable?
Suppose a questionnaire item is reworded halfway through data collection. The new wording may be clearer, but responses to the two versions may not be equivalent.
Suppose interviews become substantially longer after new topics are added. Later participants may provide evidence that earlier participants were never invited to discuss.
Suppose a laboratory instrument is replaced by a different device. Even if both measure the same construct, calibration or systematic differences may matter.
Possible responses include documenting version differences, conducting appropriate comparability checks, accounting for the change analytically, restricting particular analyses, recollecting information where ethically and practically appropriate, or acknowledging that certain comparisons are no longer defensible.
The correct response depends on the methodology and nature of the change.
Changes before main data collection are usually easier to absorb
Timing matters.
If pilot testing shows that an instruction is confusing before the main study begins, revising it may be exactly what the pilot was intended to enable.
If the same problem is discovered after half the sample has completed the procedure, the consequences are more complicated because the project now contains data generated under different conditions.
This does not mean that mid-study changes should never occur. It means that the later a change occurs, the more carefully researchers should evaluate what has already become irreversible.
This is one reason to settle consequential decisions before data collection starts wherever possible.
Changes after seeing the results require special scrutiny
The most sensitive timing occurs when researchers have already observed data relevant to the decision being changed.
Suppose a confirmatory study specifies outcome A as primary. After examining the results, researchers discover that outcome A shows little difference while outcome B produces a striking result. Redesignating B as though it had always been the primary outcome would obscure the role that observed results played in the decision.
The appropriate response is not necessarily to ignore outcome B. It may be scientifically interesting. The issue is transparency about its status.
Similarly, post hoc subgroup analyses, alternative model specifications, exclusions, transformations, or new hypotheses can be legitimate exploratory work. They should not be retrospectively represented as prospectively specified if they were developed after relevant results were known.
Prospective documentation, including protocols and preregistration where appropriate, helps distinguish planned analyses from later exploration.
Check whether ethics review or another approval is affected
For research governed by an approved human-participant protocol, researchers should not assume they can implement a modification merely because it seems scientifically sensible.
Under U.S. Department of Health and Human Services regulations, changes in approved research generally may not be initiated without Institutional Review Board review and approval except when necessary to eliminate apparent immediate hazards to participants.
NIH's Intramural Research Program similarly instructs investigators to submit modifications to approved protocols, consent forms, study tools, recruitment materials, and related documents for IRB approval before implementing the changes, apart from changes necessary to eliminate an immediate hazard.
These requirements describe particular U.S. regulatory and institutional systems. Other jurisdictions and institutions have their own processes. Follow the rules governing your study.
Watch Out
Do not implement a change to an approved human-participant study and plan to “update the ethics paperwork later.” Determine first whether the modification requires prior review, approval, notification, revised consent materials, or another institutional action under the rules that apply to your project.
Ethics amendments can themselves create new dependencies
A protocol modification may affect more than the ethics document.
If revised procedures require approval before implementation, data collection may need to pause. If new participant information is required, recruitment materials or consent forms may need revision. If another site is added, institutional authorization may also be necessary.
One change can therefore create a new dependency chain:
Problem identified Recruitment is too slow at the original site.
Proposed change Add another recruitment site.
Governance check Determine whether ethics modification and site authorization are required.
Documentation updated Revise the protocol, recruitment materials, data plan, and other affected documents.
Required approvals obtained Do not begin activities governed by the modification until authorized.
Revised procedure implemented Track which participants or observations fall under which protocol version where relevant.
The project timeline should be updated to reflect these new dependencies rather than assuming the scientific decision can be implemented immediately.
Update the protocol rather than maintaining parallel realities
A common documentation problem develops when the protocol says one thing, the ethics application says another, the questionnaire has been revised, the data-management document still reflects an older procedure, and the research team is following instructions communicated through email.
At that point, the study has several competing versions of reality.
Maintain a current master protocol or equivalent authoritative plan. When consequential changes are approved or adopted, update the affected documents consistently.
A simple version convention can help:
Protocol v1.0 · Initial approved protocol
Protocol v1.1 · Minor documented operational revision
Protocol v2.0 · Substantive approved amendment
The exact numbering system is less important than being able to identify which version was current at a particular time.
Keep a change log
Version numbers tell you that something changed. A change log tells you what and why.
For each consequential modification, consider recording:
- date;
- protocol version;
- what changed;
- why it changed;
- what evidence or event prompted the decision;
- whether relevant outcome data had already been observed;
- which documents or procedures were affected;
- whether formal review or approval was required;
- when approval was obtained where applicable;
- when the change took effect; and
- what implications it has for analysis or reporting.
This does not need to become an administrative monument. A concise table can preserve information that becomes surprisingly difficult to reconstruct months later.
Record when the change took effect
The effective date matters when data have already been collected.
Suppose questionnaire version 1 was used for participants 001–086 and version 2 for participants 087–150. That distinction may become relevant during cleaning and analysis.
The same applies to:
- changed eligibility criteria;
- new recruitment channels;
- different interview guides;
- equipment replacements;
- new data collectors;
- revised intervention procedures; or
- changes in coding or processing rules.
Without effective-date information, researchers may know that the study changed but be unable to identify which data were affected.
Preserve previous versions
Do not overwrite the old protocol, questionnaire, analysis plan, or consent form so completely that the earlier version disappears.
Previous versions may be needed to reconstruct what participants experienced, explain differences in the dataset, respond to reviewers, demonstrate compliance, or report deviations transparently.
Store superseded versions securely and label them clearly so that researchers do not accidentally continue using them.
Tell the research team what changed
A protocol amendment that nobody implementing the study knows about is not an operational amendment.
When procedures change, identify who needs to know and what they need to do differently.
This may involve:
- updated training;
- revised manuals or instructions;
- replacement forms;
- updated survey links;
- changes to database fields;
- new equipment settings;
- revised consent materials; or
- confirmation that obsolete materials have been withdrawn.
For multisite or multi-investigator studies, implementation should include verification that the correct version is being used rather than assuming that sending an email solved the problem.
Distinguish a planned amendment from a protocol deviation
These terms can have specific meanings under institutional or regulatory systems, so use the definitions applicable to your study.
Conceptually, however, there is an important difference.
Planned change
The research team decides prospectively that the protocol should be modified and follows the applicable process before implementing the new plan.
Deviation
What actually happened differs from the applicable protocol, whether because of error, circumstances, emergency, participant-specific conditions, or another cause.
A participant missing a scheduled measurement is not necessarily the same kind of event as formally changing the measurement schedule for all future participants.
The appropriate documentation, reporting, and corrective response depend on the governing system and the significance of the event.
If the change is caused by an error, address the error as well as the protocol
Suppose the research team discovers that several participants received an outdated questionnaire because an old survey link remained active.
The response should not simply be to update the protocol.
Ask:
- Which participants were affected?
- What data differ?
- Does the problem affect participant rights or safety?
- Does it require institutional reporting?
- Can the affected observations be used?
- How will the issue be handled analytically?
- Why did the obsolete link remain active?
- What process change will prevent recurrence?
Corrective action addresses the immediate problem. Preventive action addresses the system that allowed it to occur.
Reassess the sample-size or sampling implications
Changes in recruitment, attrition, eligibility, sites, outcomes, analysis, or data quality may affect sampling assumptions.
For quantitative studies, a revised design or primary analysis can alter the assumptions underlying sample-size planning. For qualitative studies, changes to population, sampling logic, or the phenomenon being investigated may alter what constitutes an adequate and coherent sample.
Do not automatically preserve the original target simply because it already appears in the protocol.
At the same time, changing sample-size targets after observing outcome results can create different inferential concerns from revising them because recruitment feasibility changed. The timing and rationale matter.
Reassess the analysis plan
A change in the data-generating process may require a corresponding change in analysis.
Adding another site may introduce clustering or site-related heterogeneity worth considering. Changing an instrument may affect score construction or comparability. Losing a measurement occasion may alter a longitudinal model. Changing sampling procedures may alter the population to which inferences can reasonably apply.
The revised analysis should respond to the revised study rather than mechanically applying a plan written for data that no longer exist in the same form.
Document which analyses were prespecified under the original plan, which were revised before relevant outcomes were observed, and which arose after examining the data where that distinction matters.
Reassess the timeline instead of moving only one date
Suppose recruitment takes four weeks longer than planned.
Do not simply move “recruitment complete” four weeks later and leave analysis, writing, supervisor review, and submission unchanged.
If those stages depend on recruitment and data collection, the delay propagates.
Recalculate the affected research project milestones and determine whether existing schedule buffer can absorb the change.
If not, an actual project assumption may need revision.
Reassess the budget and resources
Time extensions and methodological changes can have financial consequences.
Additional recruitment may increase incentives or advertising costs. Another site may create travel or coordination expenses. Extended data collection may require staff contracts to continue. A replacement instrument may require licensing. Additional laboratory processing may consume supplies.
For funded research, some changes may also require sponsor approval or formal budget modification.
Do not treat the scientific plan and resource plan as independent when one changes the other.
Reassess data management and privacy
Changes can create new data-management requirements.
Adding identifiable follow-up information changes the data being held. Adding a site may introduce data transfer. A new data source may have different use restrictions. Extending retention may affect participant information or institutional requirements. Changing software may affect where information is stored and who can access it.
Review the data-management plan whenever a change affects what data exist, where they move, who can access them, or how long they are retained.
Reassess registrations, funder requirements, and other commitments
Some studies have commitments beyond the internal protocol and ethics approval.
A project may be preregistered. A clinical trial may be registered. A funder may have approved particular aims or milestones. A registered report may have an accepted Stage 1 protocol. A data-use agreement may limit analyses to a specified research purpose.
A substantive change may therefore require updating or notifying more than one system.
The applicable requirements vary. Check the specific registry, funder, journal, sponsor, data provider, or contractual agreement rather than assuming that updating the protocol alone is sufficient.
Some changes should trigger reconsideration of whether the study should continue
Adaptation is not always the correct response.
Suppose recruitment is so poor that the intended study cannot produce useful evidence. A critical data source becomes permanently unavailable. The intervention cannot be delivered with acceptable fidelity. A newly recognized risk changes the ethical balance substantially. The remaining timeline cannot accommodate the required follow-up.
In such cases, continuing because considerable effort has already been invested can create a sunk-cost problem.
The responsible choices may include narrowing the question, redesigning the study, converting the work into a feasibility study where scientifically appropriate, postponing the project, or stopping it.
A research plan should be adaptable, but adaptation should preserve a worthwhile and defensible study rather than merely preserve the existence of the project.
Do not hide changes in the final report
If the conducted study differs consequentially from the original plan, readers may need to know.
Reporting guidelines make this particularly explicit in some research designs. CONSORT 2025 asks randomized trial reports to identify important changes to methods after trial commencement, including changes to outcomes or analyses, together with reasons. SPIRIT 2025 likewise includes protocol-amendment information among the items relevant to trial protocols.
Those guidelines concern randomized trials and should not be mechanically applied to every methodology. The broader principle is valuable: consequential methodological changes should not be concealed simply because they complicate the narrative of a perfectly executed study.
Transparency allows readers to evaluate whether and how the change affects interpretation.
Report the reason, timing, and consequence of important deviations
A useful description of a change answers several questions:
- What was originally planned?
- What was actually done?
- Why did the change occur?
- When did it occur?
- Had relevant data already been observed?
- Which participants or observations were affected?
- What analytical or interpretive consequences followed?
Not every minor logistical adjustment belongs in the published manuscript. The level of reporting should match the significance of the change and the standards of the methodology and reporting guideline being used.
Do not rewrite history
Once a study is complete, it can be tempting to edit the methods section until the entire project appears to have followed one uninterrupted plan.
That produces a cleaner story, but sometimes a less accurate research record.
If the study evolved, describe the relevant evolution. If an analysis was exploratory, call it exploratory. If recruitment changed because access failed, explain the change where it matters. If an instrument version changed, report enough information for readers to understand the data.
Scientific reporting does not require every project to have unfolded exactly as anticipated. It requires the account of what happened to be trustworthy.
Learn from the change after the project
Changes are also data about your research process.
If every project requires an emergency recruitment expansion, your recruitment assumptions may need revision. If instruments repeatedly require substantial changes after launch, your pretesting process may be too weak. If analysis plans repeatedly change because important variables were overlooked, analytical planning may need to happen earlier.
After the project, compare the original assumptions with what actually occurred.
This can improve future estimates of which research activities take longer than expected and help future plans become more realistic without becoming more rigid.
